The Divided (But Not More Predictable) Electorate: Analyzing American Presidential Vote Decisions with Machine Learning

• In Event: Modeling Voting Behavior in the 50 States

Brief Overview

We use public opinion surveys from 1952 to 2020 with machine learning to calculate out-of-sample predictions of respondents' voting decisions in presidential elections, and find that demographics do not predict vote choices better over time.

Authors

• Seo-young Silvia Kim, American University

• Jan Zilinsky, New York University